Decoupling capacitors necessary on bench power supply input on PCB?

Question

I'm planning to power my PCB directly from a 5V and 12V bench power supply. The 5V rail will power 10+ ICs (various AND/OR logic gates, high speed comparators, and a uC). I'm estimating a max of a few hundred mAs of current draw to power everything on this rail. The 12V rail is solely for the drive high power supply of a MOSFET driver IC.

I have adequate decoupling capacitors as close as possible to each IC according to their datasheets, but given the main power is coming directly from a bench power supply, will I need any decoupling capacitors close to the power input connector?

I know when working with LDOs/switching regulators, it's necessary to have a combination of bulk capacitors and ceramic capacitors to decrease impedance over a wide frequency range and reduce output voltage ripple. Intuitively, I'd think it's necessary in my case as well since the power from the supply will travel through extra connectors/pins, which adds inductance. So far, I have a 22uF bulk cap and 47nF ceramic on the 5V rail as insurance. I'd love some guidance on how I can better design this in the future. Thanks!

Answer 1

Adding ceramic and bulk capacitors on your power rails is certainly a good practice. They do very little harm(by increasing the inrush current when you poweron your circuit, and resulting in a very small increase in leakage current via the capacitors).

A well designed power supply system for circuits that demand higher reliability will often do a few tricks, adding decoupling capacitors is certainly one of those. Additionally how you route your vcc and ground track on PCB can also play vital role on reliable performance of microcontrollers(the comparators and other basic logic/analogue chips are usually a bit more forgiving). Fundamentally this is important because a microcontroller that could reset in middle of doing a complex operation sequence could have worrying consequences(but you should generally be ready for that in terms of software anyways.

Depending on how much reliability you would want to achieve, you could use try to minimize the length of tracks for vcc and ground. Use wider tracks whereever possible and try to minimize potential drop in both vcc and ground tracks. If your circuit is spread across multiple boards, it could also make sense to use dedicated buck regulators for each module feedingoff a higher voltage common vcc. If your circuit would deal with higher voltage systems, then your architecture would change even further and you would want modules to be able to feed off isolated supplies optically talking to other modules. Typically if you reach that sort of relaibility expectations, you would want to rely on high performance power supplies and build redundancies.

All that said, for a lab based experimental circuit expected to run no more than the length of the experiment, you are already good.

Answer 2

The damping factor of a series RLC circuit is:

\$ \zeta = \frac{R}{2} \sqrt\frac{C}{L} \$

A well damped circuit has \$ \zeta > 1\$.

The worst case scenario for underdamped \$ \zeta \$ is low C, high L, low R. If you have only low capacitance, low ESR caps on your board like ceramics, plus wire inductance, then the series RLC circuit formed from the bench supply's output cap, wire inductance, and board caps can have low damping factor. Combined with switching loads like digital chips or your MOSFET driver this can lead to an unhealthy amount of ringing on the local VCC.

Also, like many voltage regulators, the bench supply may be conditionally stable depending on load capacitance and ESR, and in this case low C and low ESR is also the worst case.

The solution is to add more resistance to increase \$ \zeta \$. You can't put too much resistance in series with the supply of course, so this resistance usually takes the form of the ESR of the bulk cap on the board. Value is uncritical, any general purpose cap with a few ohms ESR will do just fine.

You can try it, probe VCC with a scope while the circuits are drawing pulsed current, with and without bulk cap. If you see less ripple with the cap, that's the cap working. If you see less ringing, that's not the actual capacitance, it's the ESR giving better damping.